Industry 4.0 and digital manufacturing

Why it matters

CIM in the 1980s tried to link design, manufacturing and business through central computers; Industry 4.0 pursues the same goal with cheap sensors, networks, cloud computing and analytics. Indian manufacturers — from auto-component suppliers to process plants — are retrofitting machines with sensors to track overall equipment effectiveness (OEE), predict failures and trace every part. An engineer should be able to separate the real engineering ideas from the buzzwords and put numbers on the benefits.

Key ideas

The four industrial revolutions. Mechanisation with water and steam power (late 18th century); mass production with electricity and the assembly line (late 19th–early 20th century); automation with electronics, PLCs, CNC and IT (from the 1970s); and Industry 4.0 — cyber-physical systems connected through networks (the term was coined in Germany in 2011).

Cyber-physical systems (CPS). Physical machines with embedded sensors, computation and communication, so that the physical process and its digital representation update each other continuously. A CNC machine that reports spindle load, alarms and part counts over the network and receives programs and schedules is a simple CPS.

Enabling technologies.

• Industrial Internet of Things (IIoT) — sensors, smart devices and gateways on machines; standard protocols such as OPC UA and MQTT carry the data.
• Edge and cloud computing — time-critical processing near the machine (edge), large-scale storage and analytics in the cloud.
• Big data and analytics / AI — statistical process control, anomaly detection, machine learning for predictive maintenance and visual inspection.
• Digital twin — a virtual model of a product, machine, line or plant kept in step with its physical counterpart by live data, used to simulate, optimise and predict. (A model without the live data link is just a simulation; data flowing only one way is sometimes called a digital shadow.)
• Additive manufacturing, collaborative robots (cobots), autonomous mobile robots, augmented reality for maintenance and training.
• Horizontal and vertical integration — vertically from sensors and PLCs through SCADA and the manufacturing execution system (MES) to enterprise resource planning (ERP) (the ISA-95 levels); horizontally across suppliers, plants and customers.
• Cybersecurity — connecting operational technology to IT networks exposes machines to attack; network segmentation, access control and patching become production issues.

Design principles. Interoperability, virtualisation (digital twins), decentralised decision-making, real-time capability, service orientation and modularity.

Smart factory use cases. Real-time OEE dashboards, condition monitoring and predictive maintenance (vibration, temperature, current signatures), automated quality inspection with machine vision, traceability by part serial number, energy monitoring, and mass customisation (batch size one) with flexible cells.

Measuring the benefit. Two simple, widely used measures:

• OEE — the product of availability (running time / planned time), performance (ideal cycle time × parts made / running time) and quality (good parts / parts made). It tells you which of the three losses to attack.
• Inherent availability — from mean time between failures (MTBF) and mean time to repair (MTTR). Predictive maintenance raises MTBF (fewer surprise failures) and cuts MTTR (parts and people ready before the stop).

Limits. Connected machines do not fix a badly designed process; data without a decision attached to it is cost. Legacy machines need retrofitting, and skills in data and OT security are scarce.

Formulas

OEE = A × P × Q

• A availability = run time / planned production time; P performance = (ideal cycle time × total parts) / run time; Q quality = good parts / total parts. All dimensionless (0–1).

OEE = (good parts × ideal cycle time) / planned production time

• Equivalent shortcut.

A_i = MTBF / (MTBF + MTTR)

• Inherent availability; MTBF and MTTR in hours.

D = n · f_s · b

• Raw data rate (bytes/s); n number of sensor channels; f_s sampling frequency (Hz); b bytes per sample.

Worked examples

Example 1 (standard). In a 480 min shift a machine is down for 60 min. Its ideal cycle time is 1.0 min; it makes 380 parts, of which 361 are good. Find the OEE.

1. Run time = 480 − 60 = 420 min; A = 420/480 = 0.875.
2. P = (1.0 × 380)/420 = 0.905.
3. Q = 361/380 = 0.95.
4. OEE = A × P × Q = 0.875 × 0.905 × 0.95 = 0.752 (75.2 %). Check: 361 × 1.0/480 = 0.752 ✓.
5. The biggest loss is availability (12.5 % of time down) — that is where monitoring should start.

Example 2 (GATE level). A machine has MTBF = 200 h and MTTR = 8 h. Condition monitoring raises MTBF to 250 h and cuts MTTR to 3 h. (a) Find the change in availability and the extra productive hours in 6000 h of operation. (b) Twenty vibration channels are sampled at 10 kHz with 2 bytes per sample. Find the raw data per day.

1. A_i = MTBF / (MTBF + MTTR): before = 200/208 = 0.9615; after = 250/253 = 0.9881.
2. Gain = 0.9881 − 0.9615 = 0.0266, i.e. about 2.7 percentage points; in 6000 h that is 0.0266 × 6000 ≈ 160 h of extra availability.
3. (b) D = n · f_s · b = 20 × 10 000 × 2 = 400 000 bytes/s = 400 kB/s.
4. Per day: 400 000 × 86 400 = 3.456 × 10¹⁰ bytes ≈ 34.6 GB/day — which is why vibration data is usually reduced to features (RMS, peaks, spectra) at the edge before it is sent to the cloud.

Common mistakes

• Treating Industry 4.0 as "more automation". Its core is connectivity, data and integration; a fully manual line can still be data-driven.
• Calling any 3-D model a digital twin — a twin needs a live data link to a specific physical asset.
• Computing performance with actual cycle time instead of ideal cycle time, or multiplying percentages without converting them to fractions.
• Using MTTR + MTBF incorrectly: availability is MTBF/(MTBF + MTTR), not 1 − MTTR/MTBF.
• Ignoring cybersecurity when connecting machines to the internet.

For GATE PI

Expect conceptual MCQs on CPS, IIoT, digital twins, cloud and edge computing, the ISA-95/automation pyramid and Industry 4.0 design principles. Numericals likely use OEE and its three factors, availability from MTBF and MTTR, and simple data-rate estimates. Practise identifying which loss dominates an OEE figure.

Quick check

1. What makes a digital twin different from a simulation model?
2. Planned time 450 min, downtime 50 min, ideal cycle 0.8 min, 450 parts made, 432 good. Find OEE.
3. MTBF = 150 h, MTTR = 6 h. Find availability.
4. Name the system level between PLC/SCADA and ERP.
5. Why process vibration data at the edge?

Answers: 1. It is linked to a specific physical asset by live data. 2. A = 0.889, P = 0.90, Q = 0.96 → OEE = 0.768. 3. 150/156 = 0.962. 4. Manufacturing execution system (MES). 5. Raw high-frequency data is too large to send continuously; features are extracted locally and only results are sent.